Predicting ablation-assisted nanosecond laser fabrication of glass optical diffusers by machine learning

We apply a machine learning (ML) approach to predict the properties of glass optical diffusers fabricated by ablation-assisted nanosecond laser machining. This ‘indirect’ laser machining utilizes the interaction between the glass substrate and ablated fragments which are produced by nanosecond laser ablation of a metal plate. Therefore, the present ML models need to address these two different types of interactions. Through the proof-of-concept demonstration in the present case study, we show that fully connected neural network models, trained with a relatively small number of experimentally measured data, can reasonably predict the properties of glass optical diffusers fabricated by indirect laser machining. The results predicted by the ML models illustrate the effectiveness of combining indirect laser machining with ML models, which can be a powerful and highly flexible method for a wide range of applications.